Actigraphy, the long-term assessment of wrist movements by means of a small solid-state recorder, is widely used in a variety of human research fields, among which sheep, circadian rhythms and aging. Actigraphs assess movement with the use of accelerometers, which sense accelerations resulting from muscle force as well as accelerations due to changes in the position of the sensor in the gravitational field. In the present paper a method is described to minimise gravitational artefact in movement assessment by calculating the instantaneous acceleration vector from 3 perpendicular acceleration signals. It is shown that the power spectra of single axis acceleration signals are dominated by low-frequency components (+/- 0.25 Hz) due to gravitational artefact. Spectra of the instantaneous acceleration vector indicate that 'true' movement accelerations resulting from muscle force are present in a much wider range: from 0.25 to 11 Hz. Wrist accelerations in elderly subjects were found to be of lower amplitude and frequency as compared to young subjects. It is furthermore shown that a bandpass filter of 0.25 to 2 or 3 Hz, as has been used in commercially available actigraphs, is far from optional, and may even result in a positive bias for movement detection in the elderly. This bias may underly contradictory findings in actigraphic studies on human aging. When a bandpass filter of 0.5-11 Hz is applied to a single-axis acceleration signal, the influence of gravitational artefact and bias are minimized, and the age-related decline in activity is properly detected.